Americas Semiconductors Report Interpretation
Ahead of the September 8-11 Communacopia + Technology Conference, Goldman Sachs expects robust AI infrastructure demand, a WFE upturn lasting through at least 2028, persistent memory undersupply, an early-stage analog recovery and accelerating EDA opportunities.
Summary
Ahead of the September 8-11 Communacopia + Technology Conference, Goldman Sachs expects robust AI infrastructure demand, a WFE upturn lasting through at least 2028, persistent memory undersupply, an early-stage analog recovery and accelerating EDA opportunities.
- The conference is scheduled for September 8-11 in San Francisco and includes 32 global companies across six technology segments.
- Agentic AI, hyperscaler spending and demonstrable returns from lower token costs are expected to support AI infrastructure demand.
- Goldman expects roughly 35% WFE growth in 2026, acceleration in 2027 and a strong upturn through at least 2028.
- Estimated DRAM undersupply is 5.0%, 5.9% and 3.9% in 2026E, 2027E and 2028E.
- Estimated NAND undersupply is 4.4%, 4.6% and 3.0% in 2026E, 2027E and 2028E.
- Analog unit shipments were roughly 5% above trend as of June, but Goldman still views the recovery as being in its early stages.
- Cadence's agentic-AI opportunity is estimated at approximately $3.7bn per year by 2030 and could begin appearing as early as 2H26.
Report Interpretation
Overview
This conference preview organizes the Americas semiconductor landscape around five debates: AI compute, wafer-fabrication equipment, memory and storage, analog semiconductors and EDA software. Goldman Sachs expects broadly positive company commentary, led by AI investment and constrained semiconductor supply, while identifying financing, datacenter capacity, China-based additions, content rationalization and competitive change as the principal uncertainties.
Core views
The first debate concerns whether AI infrastructure spending can remain strong and how the competitive architecture will evolve. Goldman Sachs expects companies to sound bullish, supported by robust hyperscaler spending, agentic-AI adoption and a favorable public- and private-market funding environment. Falling token costs and tangible returns from applications such as coding are expected to provide more evidence that infrastructure investment can generate economic value. The report nevertheless identifies financing availability and physical datacenter capacity—land, power and shell—as potential constraints, with political and social considerations also capable of limiting expansion. As workloads diversify, specialized accelerators are expected to complement GPUs and CPUs by improving efficiency, freeing compute capacity and reducing power use; merchant solutions should remain dominant near term, but ASICs and other accelerators should gain share over time. Open-weight and proprietary models are expected to coexist. Nvidia is expected to maintain a bullish demand outlook after guiding to 70% revenue growth in CY27, emphasize the Rubin ramp, physical AI and its ecosystem, and address competition from custom silicon. Broadcom is expected to stress AI networking and custom XPUs; AMD should discuss MI4XX rack-level systems, ROCm and the growing role of server CPUs in agentic AI; Qualcomm should focus on datacenter diversification and edge-AI benefits across industrial, IoT, automotive and smartphones; and Credo should discuss AEC traction and the staged ramp of Zero-Flap optical transceivers, ALCs and OmniConnect/Weaver products. The second debate is the duration and composition of the wafer-fabrication-equipment upturn. Goldman expects about 35% WFE growth in 2026 and sees scope for acceleration in 2027, with the overall upturn lasting through at least 2028. The principal drivers are capacity spending in DRAM, leading-edge foundry and advanced packaging, followed by improving NAND and trailing-edge logic investment. Hyperscaler AI buildouts remain an important indirect driver, while potential spending from non-traditional customers such as Terafab and renewed Intel activity are additional debate points. The report argues that companies with the greatest deposition and etch exposure are best placed, highlighting Buy-rated Applied Materials and Lam Research; it also highlights Buy-rated Onto Innovation and Qnity for advanced packaging. Applied Materials is expected to discuss above-market growth across foundry, DRAM and advanced packaging. Lam may address share gains across DRAM, foundry and NAND and the timing of its new 55% gross-margin target. KLA should emphasize inspection, metrology and advanced packaging; Teradyne may discuss outperforming its 2H26 targets, incremental GPU share in 2027 and optical-networking and robotics opportunities over the next two years. Qnity is expected to discuss consumables, possible MSI acceleration and interconnect strength; GlobalFoundries may highlight datacenter, smartphone and emerging IP drivers; Camtek should address back-end inspection; Amkor should balance its long-term Arizona advanced-packaging expansion against near-term smartphone crosswinds; and Entegris is expected to focus on 2027 growth and margin improvement. The third debate concerns memory and storage supply, server content and valuation. Goldman expects management teams to remain upbeat about tight supply-demand conditions through 2027, consistent with second-quarter earnings commentary and Nvidia's recent report. Its estimates show DRAM undersupply of 5.0%, 5.9% and 3.9% in 2026E, 2027E and 2028E, respectively, and NAND undersupply of 4.4%, 4.6% and 3.0%. Investors are expected to test how planned additions by China-based CXMT in DRAM and YMTC in NAND could alter 2028 supply, and whether memory-content rationalization in some AI server configurations could weaken demand. Valuation also depends on the execution of capital-return plans generally ranging from 50% to 100% of excess free cash flow. Emerging technologies—including HAMR for hard drives and HBF for NAND flash—could reshape product differentiation. SanDisk is expected to reiterate its Investor Day targets, explain the effects of long-term agreements under different scenarios and discuss NAND's role in AI inference and its HBF roadmap. Seagate should provide customer feedback on Mozaic 4 qualification, the Mozaic 5 ramp and medium-term pricing and margins. Goldman expects both companies to describe continued demand growth and limited bit-supply additions over the next 18 months, retaining Buy ratings on SanDisk for tight NAND supply and improving product mix and on Seagate for tight HDD supply, strong pricing and HAMR leadership. The fourth debate is whether the analog recovery can broaden and persist. Goldman expects positive commentary on improving demand visibility, longer lead times, shipments closer to end demand and pricing actions that could more than offset higher input costs. Analog units were roughly 5% above trend as of June, but the institution views this as the beginning rather than the end of the cycle because shipments had remained below trend for much of the prior four years and demand improvement is broadening. AI datacenter demand may provide a structural extension, although the pace of recovery in traditional markets—especially automotive—remains important. NXP is expected to maintain constructive commentary on automotive secular growth without signaling customer restocking or more than a neutral pricing effect. SiTime should discuss synergies from the acquired Renesas timing portfolio, positive datacenter demand and smartphone-specific drivers. Microchip and Texas Instruments are expected to focus on datacenter exposure, distributor restocking and inventory conditions in investor meetings. Goldman is constructive on NXP for operational execution and SiTime for AI-datacenter momentum combined with solid volumes elsewhere. The fifth debate is whether custom-chip proliferation and agentic workflows can accelerate EDA and IP growth. Goldman expects Cadence to present a bullish long-term outlook, arguing that a broader set of customers designing custom chips increases demand for critical EDA tools and reusable IP. Autonomous agents could improve design productivity and create incremental monetization, producing faster growth than in the pre-AI period. The report expects supporting evidence on AI-related content, pricing and regional demand, including China. Goldman's analysis estimates that the shortage of chip-design engineers created or intensified by custom AI silicon gives EDA vendors a roughly $3.7bn annual agentic-AI opportunity by 2030. It says this opportunity is not reflected in Street estimates and could begin becoming visible as early as 2H26.
Analysis framework
Goldman Sachs begins with issues identified through investor conversations, converts them into five subsector debates and then states the commentary it expects from conference participants. It supports those expectations with recent company results and guidance, industry supply-demand estimates, shipment trends, backlog and customer forecast visibility, and company-specific product, capacity, pricing and capital-return milestones.
Methodology notes
Semiconductor capacity and supply-demand balance analysis
The report compares anticipated demand with capacity additions across WFE, DRAM, NAND, HDD and analog markets to judge the duration of tightness, the industry upturn and pricing support.
Cycle-stage assessment using shipments, backlog and spending visibility
Goldman uses below-trend historical analog shipments, the recent move to roughly 5% above trend, equipment backlogs and extended customer forecasts to determine whether recoveries are beginning, maturing or likely to persist.
AI-infrastructure demand transmission through the semiconductor ecosystem
The report traces hyperscaler and agentic-AI demand through datacenter capacity, GPUs, CPUs, ASICs, networking, equipment, advanced packaging, memory and EDA software to identify where spending and constraints affect companies.
Conference expectation mapping
The preview identifies current investor debates and specifies what management teams are expected to say, making the conference a test of whether company commentary confirms or challenges prevailing expectations.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Nvidia (NVDA)A central beneficiary and demand indicator for AI infrastructure spending.
- Strengths
- Expected Rubin ramp, ecosystem strength, physical-AI opportunity and recent guidance for 70% CY27 revenue growth.
- Weaknesses
- Must address increasing competition from custom silicon and specialized accelerators.
- Comparison
- Merchant GPU solutions are expected to remain dominant near term even as ASICs gain traction.
- Risks
- Datacenter capacity constraints and moderation in AI capital spending could affect the demand trajectory.
- BroadcomLinked to AI networking and custom-XPU growth.
- Strengths
- Leadership in AI networking and custom XPUs, with visibility across multiple XPU customers.
- Comparison
- Positioned on the custom-silicon side of the merchant-versus-custom architecture debate.
- Risks
- Competitive changes within custom silicon could affect its positioning.
- AMD (AMD)Competes in AI accelerators and benefits from the rising importance of server CPUs.
- Strengths
- Growing datacenter presence, MI4XX rack-level solutions, ROCm software and server-CPU exposure to agentic AI.
- Weaknesses
- The company still needs to improve its competitive position.
- Comparison
- Presented as an alternative in a market currently led by merchant GPU solutions.
- Risks
- Competition across GPUs, CPUs and custom accelerators.
- Applied MaterialsBuy-rated beneficiary of the WFE upturn.
- Strengths
- Exposure to foundry, DRAM and advanced packaging, with potential to outgrow the broader market.
- Comparison
- Highlighted with Lam Research among companies best exposed to deposition and etch spending.
- Risks
- The benefit depends on the duration and composition of customer capital spending.
- Lam ResearchBuy-rated beneficiary of equipment spending and potential market-share expansion.
- Strengths
- Exposure to DRAM, foundry and NAND and a new 55% gross-margin target.
- Weaknesses
- The timing of reaching the margin target remains a discussion point.
- Comparison
- Highlighted with Applied Materials for strong deposition and etch exposure.
- Risks
- A shorter WFE cycle or weaker memory spending would reduce the expected benefit.
- Onto InnovationBuy-rated beneficiary of secular advanced-packaging growth.
- Strengths
- Positioning in an area Goldman expects to show continued momentum.
- Comparison
- Highlighted alongside Qnity for exposure to advanced packaging.
- QnityBuy-rated beneficiary of advanced packaging, consumables and interconnect demand.
- Strengths
- Potential MSI acceleration and incremental interconnect strength in 2027.
- Comparison
- Highlighted alongside Onto Innovation for advanced-packaging exposure.
- SanDisk (SNDK)Buy-rated beneficiary of constrained NAND supply and AI-inference storage demand.
- Strengths
- Limited near-term NAND additions, improving portfolio and mix, long-term agreements and an HBF roadmap.
- Comparison
- Paired with Seagate as a storage beneficiary of limited supply growth, but exposed to NAND rather than HDD.
- Risks
- China-based NAND additions and content-per-box rationalization could weaken supply-demand conditions.
- SeagateBuy-rated beneficiary of tight HDD supply and pricing.
- Strengths
- HAMR technology leadership, strong pricing and the Mozaic 4 and Mozaic 5 product ramps.
- Comparison
- Paired with SanDisk as a storage beneficiary, with differentiation based on HDD and HAMR technology.
- Risks
- Qualification and product-ramp execution could influence medium-term performance.
- NXP (NXPI)Analog semiconductor exposure to automotive secular growth and the broader recovery.
- Strengths
- Operational execution on key metrics and constructive automotive demand trends.
- Weaknesses
- No expected signal of customer restocking, while pricing is expected to have only a neutral effect.
- Comparison
- One of Goldman's two constructive analog names, alongside SiTime.
- Risks
- A delayed recovery in automotive and other traditional end markets.
- SiTime (SITM)Analog and timing beneficiary of AI datacenter demand.
- Strengths
- AI-datacenter momentum, solid volumes elsewhere and potential synergies from the acquired Renesas timing portfolio.
- Comparison
- Goldman's constructive view reflects stronger idiosyncratic growth than the broader analog cycle alone.
- Risks
- Execution on acquisition synergies and smartphone-specific demand.
- CadenceEDA and IP beneficiary of custom-chip proliferation and agentic-AI workflows.
- Strengths
- Critical design tools, reusable IP and autonomous agents; Goldman estimates an incremental opportunity of approximately $3.7bn per year by 2030.
- Weaknesses
- The incremental opportunity is expected to emerge over time rather than immediately.
- Comparison
- EDA vendors are presented as uniquely positioned to monetize the shortage of chip-design engineers.
- Risks
- The pace of customer adoption and monetization of agentic workflows may differ from expectations.
Key data
- Conference datesSeptember 8-11, 2026Communacopia + Technology Conference in San Francisco
- Participating companies32 global companiesAcross Digital/AI Semis, EDA Software, Analog, SemiCap Equipment, Memory/Storage and IT Services
- Nvidia CY27 revenue guidance70% growthRecent guidance cited as context for an expected bullish demand outlook
- 2026 WFE growth estimate~35%Goldman sees potential for growth to accelerate further in 2027
- WFE upturn durationThrough at least 2028Driven principally by DRAM, leading-edge foundry and advanced packaging
- Lam Research gross-margin target55%The expected timing of achievement is a conference discussion point
- DRAM supply-demand deficit5.0% / 5.9% / 3.9%Estimated undersupply for 2026E / 2027E / 2028E
- NAND supply-demand deficit4.4% / 4.6% / 3.0%Estimated undersupply for 2026E / 2027E / 2028E
- Memory capital-return plans50%-100% of excess free cash flowTypical range cited; execution may influence valuation
- Expected period of limited memory bit-supply additionsNext 18 monthsExpected message from SanDisk and Seagate
- Analog shipments versus trendRoughly 5% above trendAs of June, following below-trend shipments over much of the prior four years
- Cadence agentic-AI opportunity~$3.7bn/year by 2030Goldman says it is absent from Street estimates and may become evident as early as 2H26
Impact & implications
The report's central implication is that AI spending is supporting multiple layers of the semiconductor ecosystem rather than only GPUs. Goldman sees deposition and etch vendors and advanced-packaging suppliers as especially exposed to the WFE upturn, expects constrained memory supply and capital returns to support memory valuations, and views AI datacenters as an extension of the analog recovery. Custom silicon also increases the need for EDA tools and IP, while agentic workflows may create an additional monetization opportunity for Cadence.
Risks
- Public- or private-market financing may become less available for continued AI infrastructure investment.
- Land, power and datacenter-shell availability, including political and social constraints, could limit AI capacity growth.
- AI capital-expenditure durability remains sensitive to the macroeconomic and interest-rate backdrop.
- ASICs and specialized accelerators could intensify competition with merchant GPU and CPU suppliers.
- China-based DRAM and NAND capacity additions could loosen supply conditions from 2028.
- Memory-content rationalization in some AI server configurations could reduce content-per-box demand.
- A delayed automotive or broader traditional-market recovery could constrain analog growth.
- Near-term smartphone crosswinds could affect Amkor and other exposed suppliers.
What to watch
- Management commentary on hyperscaler spending, agentic-AI returns and public- and private-market financing.
- Updates on datacenter land, power and shell availability heading into 2027.
- The competitive balance among GPUs, CPUs, ASICs and other specialized accelerators.
- Whether 2027 WFE growth can accelerate beyond the approximately 35% expected for 2026 and persist beyond 2028.
- Relative equipment spending across leading-edge foundry, DRAM, NAND, advanced packaging and trailing-edge logic.
- The scale and timing of CXMT and YMTC capacity additions and any AI-server memory-content reductions.
- Execution of memory-sector capital returns equal to 50%-100% of excess free cash flow.
- Analog lead times, distributor restocking, automotive demand and the revenue and margin effect of pricing actions.
- Cadence evidence on AI-related content, pricing and autonomous-agent monetization, potentially beginning in 2H26.
- Product milestones including Nvidia Rubin, SanDisk HBF, Seagate Mozaic 4 and 5, and HAMR adoption.